Journal of Fluids and Structures (2000) 14, 323}338

doi:10.1006/j#s.1999.0274 available online at http://www.idealibrary.com on

FLOW-INDUCED MULTIPLE-MODE VIBRATIONS

OF GATES WITH SUBMERGED DISCHARGE
P. BILLETERR
Laboratory of Hydraulics (VAW), Swiss Federal Institute of Technology
CH-8092 Zurich, Switzerland
AND

T. STAUBLI
Fluidmechanics and Hydromachines, Zentralschweizerisches echnikum uzern
CH-6048 Horw, Switzerland

(Received 10 July 1997, and in "nal form 18 October 1999)

An experimental investigation of #ow-induced vibrations of gates with multiple degrees of

freedom is presented. An under#own vertical gate plate with submerged discharge was allowed
to oscillate both in the cross-#ow (z-) and in the streamwise (x-) direction. The two purposes of
the investigation were to further the insight into the hydrodynamic coupling mechanisms of the
two vibration modes and to determine the interaction of the unsteady lift and drag forces.
Self-excited vibration tests were run with reduced velocities < and < from 0)8 to 14, covering
PX PV
a range in which the instability-induced excitation (IIE) due to impinging-leading-edge vortices
(ILEV) as well as the transition to galloping (MIE) occurred. The ratio of the natural
frequencies of the two vibration modes f /f , the gate opening ratio s/d, and the submergence
V X
of the gate plate were varied. Depending on the ranges of reduced velocities and frequency
ratios, a complex interaction of two di!erent kinds of instability-induced excitation was
detected. Furthermore, it was found that streamwise IIE-excitation and cross-#ow galloping
coexist. To assess the relevant #uid dynamic ampli"cation and attenuation mechanisms,
simultaneous body response and #ow velocity measurements were carried out.
 2000 Academic Press

1. INTRODUCTION
TYPICAL FEATURES OF FLOW-INDUCED VIBRATIONS of under#own gates are excitation due to
instability or lateral de#ection of a single shear layer, in contrast to the #ow around a blu!
body where two shear layers might interact. There is a strong mean-pressure gradient in the
streamwise direction. For small gate openings, minor variations of the gate position can
dramatically alter both the discharge beneath the gate and the pressure gradient between
the head- and the tailwater. Under#own gates with one degree of freedom (d.o.f.) were
investigated among many others by Thang (1984, 1990), Thang & Naudascher (1986),
Hardwick (1974) and Kolkman (1976), concerning the excitation of cross-#ow (vertical)
vibrations, and by Jongeling (1988), Thang (1990) and Ishii & Knisely (1989, 1990) with
respect to streamwise (horizontal) vibrations. A variety of excitation mechanisms was found
[see, e.g., Naudascher & Rockwell (1994)].
For gates with a rectangular gate-lip geometry and #ow separation at the leading
edge (Figure 1), impinging-leading-edge-vortex excitation (ILEV) and movement-induced

R Current Address: GRUNER Engineering Ltd., CH-4020 Basel, Switzerland.

0889}9746/00/040323#16 $35.00/0  2000 Academic Press

324 P. BILLETER AND T. STAUBLI

Figure 1. De"nition and schematic diagram of the experimental set-up.

excitation (MIE) are the most signi"cant. The ILEV-excitation results from the instability of
the e%ux shear layer, which is ampli"ed in the presence of the trailing edge of the gate lip.
The galloping-type MIE is caused by motion-dependent hydrodynamic mean forces and
in#uenced by discharge and pressure variations (Kanne et al. 1991). The ILEV-excitation
occurs for small gate openings s/d+0)5}1 and speci"c < and < , respectively. The onset
PX PV
of MIE starts for all s/d'0)5, as soon as a critical reduced velocity < is exceeded.
PX
Prototype gates, however, must be considered as systems with multiple d.o.f. Most often,
there is a variety of bending and torsion modes with typical amplitudes in the streamwise
direction. Additionally, there is a mass-oscillation mode in the cross-#ow direction for
which the whole gate acts as the oscillating mass and the restoring forces are provided by
the gate suspensions. The natural frequencies of the bending modes typically lie between
5 and 50 Hz, while the natural frequencies of the mass-oscillation modes range from 2 to
20 Hz. An interaction of several vibration modes is therefore likely to occur (Ishii & Knisely
1992; Billeter 1995). Consequently, the investigation presented aimed subsequently at
a better understanding of the hydrodynamic coupling mechanisms between two perpen-
dicular vibration modes.

2. EXPERIMENTAL SET-UP
The test facility consisted of a vertical gate plate of 1 m height which was positioned in
a horizontal laboratory #ume of about 10 m length and 1 m width (turbulence level of the
undisturbed approach #ow: Tu+2%). For the vertical direction of motion, the gate plate
was guided by linear roller bearings and suspended by coiled springs. The construction
itself, providing the vertical d.o.f., acted as a pendulum that was movable in the streamwise
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 325

direction and suspended by leaf-springs. Additional damping was supplied by electromag-

netic dampers for both directions of motion. The gate behaved as a linear mass oscillator
with two perpendicular d.o.f. It was allowed to oscillate in the cross-#ow (vertical, z) and the
streamwise (horizontal, x) directions (Figure 1). The gate was under#own and had a rectan-
gular bottom lip shape in order to produce maximum #uid dynamic excitation and to
exclude any structural coupling. The width of both the gate lip and the gate plate was
d"10 cm. The test facility simulated a strip in the mid-section of a weir gate which can be
lifted in the vertical direction and is subjected to #exural de#ections in the streamwise
direction.
The gate lip was equipped with a total of ten miniature pressure transducers which were
#ush-mounted to the structural surface. The position of the pressure transducers in the mid
section of the gate can be seen in Figure 1. The response of the body oscillator was measured
by load- and de#ection cells which were located at or close to the gate suspensions. The
signals of all the gauges were ampli"ed by two carrier-frequency ampli"ers and simulta-
neously digitized (HBM DMC 9012A, with built-in "lters). The data were collected by
a Macintosh computer using an IEEE-488 interface. The record length was typically
15;2 samples per gauge and the sampling rate was 150 Hz. Applying appropriate
FFT-techniques (overlapping Hamming windows, blockwise averaging and both frequency
and peak power estimation routines), a frequency resolution of about 0)02 Hz and an
amplitude resolution better than 0)2 dB was achieved in the frequency domain.
In order to measure the local #ow velocities and to investigate the #ow "eld around the
gate lip and in the tailwater of the gate, a Dantec-BSA two-component "bre-optic LDA-
system was used. For the assessment of the #uid dynamic excitation mechanism, both
simultaneous and synchronized measurements of the body response, the surface pressure
and the #ow velocity were carried out.

3. ANALYSIS AND PARAMETER VARIATION

Structural coupling between the two d.o.f. can be neglected provided the beams that
comprise the pendulum for the horizontal d.o.f. are su$ciently long and of su$ciently small
inertia. Upon making this assumption, the equations of motion become
zK #2b u zR #u z"F /m , (1a)
X X X UX X
xK #2b u xR #u x"F /m , (1b)
V V V UV V
where m is the oscillation mass, b "b /(2m u ) the damping ratio, u "2n f "(c /m ),
G G G G G G G G G
f the natural frequency in air, b the viscous damping factor, and c the spring rigidity
G G G
(index i stands for either z or x; Figure 1). For the #uid dynamic excitation forces F , which
UG
are nonlinear functions F(x, xR , z, zR ), the following linear approach was chosen (u "

(2gDh ) and B is the spanwise width):

u u
F "  oBdC and F "  oBdC ,
UX 2 $X UV 2 $V

with
C "C sin(u t#
)#C sin(u t#
), (2a)
$X $XX X XX $XV V XV
C "C sin(u t#
)#C sin(u t#
), (2b)
$V $VX X VX $VV V VV

being the phase shift between de#ection and excitation force. Energy is transferred from
GH
the #ow to the body if 0(
(1803. All nonlinear components of the self-excited system
GH
326 P. BILLETER AND T. STAUBLI

are included in the force coe$cients C and the phases

. C and
therefore depend
$GH GH $GH GH
on the geometric, hydraulic and structural parameters. With equation (2), the solution of
equation (1) becomes
z(t)"z sin(u t)#z sin(u t#u ), (3a)
X X V V XV
x(t)"x sin(u t!u )#x sin(u t). (3b)
X X VX V V
Since the amplitudes i and the phases u were measured, the force coe$cients C and
H GH $GH
the phases
(
Ou ) could be determined. A point located on the gate undergoing an
GH GH GH
oscillation with z"z sin(u t) and x"x sin(u t!u ) has an elliptic trajectory. Refer-
X X X X VX
ring to a coordinate system as shown in Figure 1, the gate acts as a press-shut device if the
shorter axis of the ellipse has a positive inclination a, and as a press-open device if the
inclination a is negative.
During the investigation, the following parameters were varied: (a) the gate opening s/d,
the discharge q underneath the gate and the gate submergence C "(H #H !2s)/d (the
Q M S
mean velocity u at the vena contracta is determined by s/d, q, C ); (b) the natural frequencies
 Q
and the damping ratios of the two vibration modes ( f and f were varied between 2)5 and
X V
7)0 Hz).
Self-excited vibration tests were run with reduced velocities < "u /( f d) from 0)8 to 14,
PG  G
covering a range inside which both the ILEV-excitation and the transition to MIE
(galloping) occurs ( f is the vibration frequency). The frequency ratio f /f ranged between
G V X
0)99 and 3)50. With damping ratios b between 1)5 and 3%, the Scruton numbers
G
Sc "4nm b odB were between 1)6 and 4)0. The gate opening s/d was varied between 0)5
G G G
and 1, and C ranged from 1)5 to 6. The Reynolds number Re"u d/l varied between
Q 
6;10 and 3;10.

4. RESULTS
Before discussing selected results of the investigation in detail, some overall features of
#ow-induced 2-d.o.f. vibrations will be pointed out. Figure 2 shows the power spectra of the
vertical suspension force F "z c for di!erent #ow velocities u . The frequency ratio f /f
X X X  V X
is 2)89. It can be seen that with increasing u , the dominant vibration amplitudes lie within

di!erent frequency ranges. In range A, close to the natural frequency f of the vertical d.o.f.,
X
dominant vertical vibrations due to ILEV-excitation occur (2(< (4). Range B is linked
PX
to the natural frequency f of the horizontal d.o.f. and the response of the oscillator
V
becomes much stronger in the horizontal direction (3)5(< (7). For high velocities
PV
u (< '10), the oscillation in the vertical direction is dominant again and the vibration
 PX
frequencies are within range C which is about 0)8 f . These vibrations are caused by MIE as
X
discussed in Section 1. Depending on the reduced velocities < and < , the system selects
PX PV
an appropriate mode and the properties of these modes will be discussed subsequently.
Within ranges A and B, the de#ection amplitudes were periodic and nearly harmonic. The
in#uences of both the submergence ratio C and of the gate opening s/d will not be
Q
considered in this paper. Although they a!ect quantitatively the vibration phenomena, they
do not alter the basic physical properties of the excitation mechanism.

4.1. DOMINANT VERTICAL RESPONSE, RANGE A

In Figure 3, the amplitudes z /d and x /d are plotted versus the reduced velocity < for
X X PX
di!erent frequency ratios f /f "1)01 and 1)47 and for the gate with the horizontal d.o.f.
V X
blocked (s/d+0)6). The vertical amplitudes z /d are always much stronger than the
X
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 327

Figure 2. Power spectra of the vertical suspension force F . From bottom to top, the #ow velocity u is
X 
increased. 2-d.o.f., f /f "2)89, s/d"0)8}0)9.
V X

Figure 3. Amplitudes z /d and x /d: (z : , x : ) 2-d.o.f., f /f "1)01; (z : , x : ) 2-d.o.f., f /f "1)47; (;)

X X V X V X
1-d.o.f., x-de#ection blocked.

horizontal ones (x /d). Figure 4 shows the coe$cients C and phases

for f /f "1)01
X $XX XX V X
and for the gate with only one z-d.o.f. Typically, the end of the excitation range is associated
with a sudden increase of the phase
. Comparing the z -amplitudes for f /f "1)47
XX X V X
with the data measured for a gate with the horizontal d.o.f. blocked, practically no di!erence
can be found (Figure 3). In principle, this "nding also holds for frequency ratios
f /f '1)5, although for a frequency ratio f /f +2, the vertical amplitudes might be
V X V X
328 P. BILLETER AND T. STAUBLI

Figure 4. Force coe$cient C and phase

: (C : ,
: ) 2-d.o.f., f /f "1)01; (C : ,
: ) 1-d.o.f.,
$XX XX $XX XX V X $XX XX
x-de#ection blocked.

slightly attenuated due to energy transfer from the vertical to the horizontal response (the
mechanism is not completely understood, yet). For 1 z-d.o.f. oscillations with comparable
Scruton numbers Sc and gate openings s/d, both the amplitudes z /d and the < -range,
X X PX
where the excitation occurs, agree well with the results found by Thang (1984, 1990) and
Hardwick (1974). The excitation mechanism is of a vertical ILEV-type (z-ILEV), as will be
pointed out later.
If f /f +1)0, the vertical amplitudes z /d become much larger than the amplitudes of
V X X
the gate with only one vertical d.o.f., and the range of excitation is extended (Figures 3
and 4). As mentioned before, the trajectory of the gate motion is an ellipse with the longer
axis close to both the direction and size of the amplitude z /d. Figure 5 shows the angle
X
a between the horizontal axis x and the shorter axis of the ellipse. For f /f +1, a is
X V X X
smaller than 0 and the gate acts as a press-open device. On the other hand, for f /f 51)5,
V X
a is positive and the gate acts as a press-shut device. Further explanations of the speci"c
X
behaviour of the vibration orbits will be given in Section 4.3.

4.2. DOMINANT HORIZONTAL RESPONSE, RANGE B

Streamwise vibrations of gate plates are strongly a!ected by both added mass and wave
e!ects caused by the headwater and tailwater. The natural frequency of the submerged gate
decreases roughly proportionately to (1/C). Figure 6 shows the horizontal amplitudes x /d
Q V
plotted versus the reduced velocity < for three di!erent frequency ratios f /f "1)47,
PV V X
2)07 and 2)46 as well as for the gate with the vertical d.o.f. blocked (s/d+0)6). Streamwise
vibrations within the range of 1)5(< (2 were detected as well but will not be considered
PV
in this paper. Both the amplitudes x /d and excitation range of the gate with one horizontal
V
d.o.f. are in fair agreement with results by Thang (1991) and Jongeling (1988). For the
di!erent operation states with 2-d.o.f., the maximum amplitudes x /d are not ampli"ed
V
compared to the 1-d.o.f. state. The < -range of the maximum excitation, however, is shifted
PV
toward smaller < . Again, the trajectory of the gate motion is an ellipse, but compared to
PV
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 329

Figure 5. Rotation angle a of the elliptic trajectory: open signs: a of dominant z (t), solid signs: a of dominant
X X V
x (t). (;) f /f "1)01; (, ) f /f "1)47; (, ) f /f "2)07; (, ) f /f "2)46.
V V X V X V X V X

Figure 6. Horizontal amplitudes x /d. 1-d.o.f.: (;), no z-de#ection. 2-d.o.f.: () f /f "2)46; () f /f "2)07;
V V X V X
() f /f "1)47.
V X

the ILEV-response in the vertical direction, the nondominant amplitude (z /d) is much
V
closer to the dominant one (x /d). As shown in Figure 5, the angle a between the x-axis and
V V
the shorter axis of the ellipse is always a '0. Thus, the gate behaves as a press-shut device
V
independently of the frequency ratio. The excitation mechanism causing dominant horizon-
tal response will be discussed in the following section.
330 P. BILLETER AND T. STAUBLI

4.3. EXCITATION MECHANISM

To assess the above-mentioned excitation mechanisms, force, pressure and velocity
measurements were carried out at the "xed gate. Using long time-series and averaging
techniques in the frequency domain, weak but signi"cant periodic components could be
detected in the measured signals. These components emerge from coherent #ow structures
in the shear layer separating at the leading edge of the gate (the weakly organized vortices
are caused by the Kelvin}Helmholtz instability between the separated #ow and the jet
underneath the gate). Figure 7 shows the power spectra of the gate lift force F measured for
X
di!erent #ow velocities u ; (F is equivalent to the integrated instantaneous surface pres-
 X
sure). With increasing #ow velocities, the location of the weak peaks in the spectra tend
towards higher frequencies, suggesting the existence of a convective instabilities with
speci"c Strouhal numbers Sh "f d/u . The data in Figure 7 yields to Sh +0)4, Sh +0)2,
G G   
Sh +0)1 and accordingly < +2)5, < +5, < +10.
 P P P
In Figure 8, the cross-correlation of simultaneous pressure measurements at the gate and
velocity measurements in the tailwater are plotted for increasing streamwise distance
between the gate and the location of the velocity measurement. The convection of the
persistent #ow structure is clearly visible; the convection velocity of these structures
amounts to u +0)6u . Considering the cross-correlation of 2-D velocity measurements in
A 
tailwater (Figure 9), it is observable that with increasing distance from the trailing edge of
the gate, #ow structures with successively large periods arise. The Strouhal numbers are
Sh +0)36 and Sh +0)18 which is in fair agreement with the Strouhal numbers calculated
 
from the force measurements at the rigid gate. Although the spectral peaks and the maxima
of the cross-correlations are rather weak (R +0)05}0)1), it was concluded that the
GH 
shear layer has a tendency to form coherent vortical structures; (as a consequence of the
great number of blocks used for averging the power spectra and the correlation functions,
all peaks are statistically signi"cant; see Section 2). These vortices undergo a vortex pairing
process. The pressure variations caused by this process can be detected at the gate itself.

Figure 7. Fixed gate: power spectra of the suspension force F . From bottom to top,the velocity u increases
X 
from 1)3 to 2)3 m/s.
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 331

Figure 8. Cross-correlation between cross-#ow velocity v and gate pressure p . From bottom to top, the

distance x/d between the gate and the velocity measurement is increased (the distance above the channel bed was
kept constant at z/d"!0)2).

Figure 9. Cross-correlation between the cross-#ow and the streamwise velocity (v, u). From bottom to top, the
distance x/d between the gate and the velocity measurement is increased (the distance above the channel bed was
kept constant at z/d"!0)2).

For the further analysis of the excitation mechanisms, the #uctuating pressure coe$cients
C (0) of the dominant vibration frequencies u were computed from the pressure measure-
NL H
ments along the gate lip. Assuming p (t)"p sin(u t!u ), C (0) can be calculated with
L L H HL NL
respect to the gate position z (t)"0 or x (t)"0 when u t"0, 2n, 2:
H H H
2p
C (0)" L sin (!u ), (4)
NL ou HL

332 P. BILLETER AND T. STAUBLI

where p is the pressure amplitude, and u the phase between de#ection and pressure at the
L HL
respective position (the de#ection is either z (t)"z sin(u t) or x (t)"x sin(u t) with j"z
H H H H H H
or x, depending on the vibration mode to be analysed). A positive C (0) indicates that
NL
energy is transferred from the #ow to the body and vice versa. If the #uctuating pressure is
assumed to propagate as a coherent structure along the #ow boundary, the convection
velocity u can be calculated according to
A
u "u Dx/Du , (5)
A H LK
where u is the circular frequency of the respective vibration mode, Du the phase shift
H LK
between the pressure signals measured at two di!erent positions n, m along the gate lip, and
Dx the distance between the two measurement positions n, m.
Subsequently, some properties of the ILEV-excitation causing cross-#ow gate vibrations
(range A) will be pointed out. The pressure distribution has a wavy shape with growing
amplitude towards the trailing edge as was reported previously by Thang (1991) and
Hardwick (1974). With increasing < , the wave becomes distorted and this results in an
PX
increasing instantaneous pressure gradient between the mid-lip position and the trailing
edge [Figure 10(a)]. Likewise, at the trailing portion of the gate lip, the relative convection
velocity u /u of the &&pressure wave'' (Figure 11) decreases with increasing < .
A  PX
Both of these "ndings can be considered as intrinsic features of the ILEV-excitation, if
we assume that the instability due to the shear-layer impingement is con"ned between
the leading and the trailing edge of the gate. The distortion of the #uctuating pressure
distribution and the reduction of the convection velocity are considered to be responsible
for the onset of higher harmonics and the nonlinear behaviour of the vibration mechanism

Figure 10. Instantaneous pressure distribution along the gate lip. The pressure measurement points at the
vertical faces of the gate are projected on the horizontal axis. (a) Dominant vertical vibration at u t"0 and
X
z (t)"0, < "2)7 (), 3)4 (), 3)8 (); (b) dominant horizontal vibration at u t"0 and x (t)"0. < "3)8 (),
X PX V V PV
4)3 (), 4)6 ().
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 333

Figure 11. Relative convection velocity of the pressure #uctuations. z (t) dominant: () u J*u(p !p ), ()
X A  
u J*u(p !p ); x (t) dominant: () u J*u(p !p ), () u J*u(p !p ).
A   V A   A  

towards the end of the excitation range. Supposedly, the abrupt end of the ILEV-excitation
range in Figure 3 is associated with the constant convection velocity along the gate lip
(Figure 11).
The excitation of streamwise vibration was not completely understood yet. While for
vibrations within the range of 1)5(< (2 a higher mode of ILEV-excitation can be
PV
assumed with good reason (Billeter 1998), the excitation between < "4 and 8 needs
PV
further clari"cation. According to Thang (1991), the mechanism is a &&submode'' of ILEV-
excitation. Jongeling (1988) gives an explanation that comes close to what Naudascher
& Rockwell (1994) refer to as &&wake breathing''. If we consider the instantaneous pressure
distribution along the gate lip [Figure 10(b)], we "nd that it becomes #attened with
increasing < . The relative convection velocity u /u (Figure 11) increases with increasing
PV A 
< , indicating that the vortical structures in the shear layers will be stretched towards
PV
the upper end of the vibration range (contrary to the vertical ILEV-excitation for which the
vortices are &&squeezed''). Taking the "ndings for the free shear layer beneath the "xed gate
into account, as well as the range of reduced velocities < where streamwise vibration
PV
occur, it can be concluded that a #exible gate (i.e., a #exible locus of #ow separation) will
enhance and trigger the vortical organization in the shear layer within this distinct range of
reduced velocities. Strong interaction of the shear layer and the trailing edge of the gate no
longer occurs [compare Figure 10(b) with 10(a) and Figure 12(c,d) with 12(a,b)]. The
feedback mechanism, involving the shear-layer instability and its control by the gate
motion, will be persistent, as long as the induced pressure #uctuations acting dominantly on
the downstream face of the gate [positive C (0) and C (0) in Figures 12(c) and 12(d)] are
N N
leading the gate motion. Since the excitation is caused by the motion-induced instability of
the shear layer de#ecting from the leading edge of the gate, it is suggested that the excitation
mechanism should be considered as streamwise body-resonant leading-edge vortex shed-
ding (BR-LEVS). Furthermore, it can be supposed that these "ndings could apply to other
vibration problems involving one single shear layer or two shear layers with symmetrical
vortical structures (e.g., streamwise vibrations of short rectangular cylinders in parallel
#ow).
 334
P. BILLETER AND T. STAUBLI

Figure 12. Pressure coe$cients C (0) at z (t)"0 and x (t)"0. z (t) dominant: (a) 2-d.o.f., f /f "1)47; (b) 2-d.o.f., f /f "1)01. x (t)
NL X V X V X V X V
dominant: (c) 1-d.o.f., z (t) blocked; (d) 2-d.o.f., f /f "2)5. (;)"C (0), ()"C (0), ()"C (0), ()"C (0), ()"C (0), ()"C (0).
X V X N N N N N N
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 335

Finally, the coupling between vertical and horizontal vibrations of the gate with 2-d.o.f.
are discussed. For this reason, the pressure coe$cients as de"ned in equation (4) are plotted
versus the reduced velocities. The C (0), C (0) and C (0) refer to the vertical motion of
N N N
the gate, while C (0), C (0) and C (0) are related to the streamwise motion (Figure 1).
N N N
Figure 12(a) shows typical pressure coe$cients for an ILEV excitation with f /f 51)5.
V X
The dominant #uid excitation occurs towards the trailing edge of the gate lip (C (0)). The
N
pressure coe$cients referring to streamwise excitation are negative, except for C (0) on the
N
upstream face of the gate. Since the phase shift between x (t) and z (t) is about $n,
X X
a high-amplitude z (t) motion in the upward direction causes a small-amplitude x (t)
X X
motion against the streamwise direction. The gate acts as a press-shut device with usually
clockwise rotation (at the upper end of the excitation range the rotation may change to
counterclockwise). The streamwise vibrations are excited by upstream pressure #uctuations
which are caused by discharge variations related to the #ow structures travelling along the
gate bottom, as suggested by Ishii & Knisely (1989, 1990).
In Figure 12(b), the pressure coe$cients for ILEV-excitation and a frequency ratio
f /f +1 are shown. The pressure coe$cients indicating excitation in the vertical direction
V X
behave very similar to the mechanism mentioned above, for up to < +3)2. From < +3)2
PX PX
to 3)8, C (0) increases further and C (0) decreases below zero. This behaviour can be
N N
related to the onset of the streamwise, body-resonant LEVS as pointed out later [Fig-
ure 12(d)]. The pressure coe$cients indicating streamwise ampli"cation are positive at the
tailwater and negative at the headwater face of the gate. Therefore, the ampli"cation of the
streamwise motion is caused by the pressure variations from the enhanced shear-layer
instability. It can be concluded that for f /f +1, the extension of the vibration range and
V X
the increased vibration amplitudes result from the coexistence of vertical ILEV and
streamwise, body-resonant LEVS. The x -motion is heading the z -motion by about n/2 (see
X X
Figure 5) and the gate behaves as a press-open device with counterclockwise rotation.
Figures 12(c) and 12(d) show the pressure coe$cients in the range of stream-
wise vibrations with "xed and released vertical d.o.f. It can be seen that all the pressure
coe$cients related to streamwise excitation (C (0), C (0) and C (0)) are positive. The
N N N
body-resonant shear-layer instability causes both pressure #uctuations at the trailing
portion of the gate and simultaneous pressure alterations in the head water, the latter being
produced by discharge variations related to the #uctuating shear layer.
As for the gate with 2-d.o.f., the pressure coe$cient representing vertical excitation is
reversed compared to the 1-d.o.f. situation. The vertical motion is ampli"ed by the pressure
#uctuations caused by the shear-layer instability. These #uctuations a!ect both the lower
portion of the tailwater face of the gate plate and the trailing end of the gate lip. Since
the amplitude of z (t) is smaller, the amplitude of x (t) and z (t) lags x (t) by a phase
V V V V
angle smaller than n, and the gate will act as a press-shut device again; but, compared to
the dominant ILEV-excitation, the rotation is counterclockwise.

4.4. TRANSITION TO MIE

Figure 13 shows the vertical vibration amplitudes z /d for the gate with 2-d.o.f. and
X
frequency ratios between f /f "2)07 and 3)50 (s/d+0)8). It seems that neither the onset
V X
reduced velocity < nor the vibration amplitudes of the MIE-mechanism are severely
PX 
a!ected by the frequency ratio. As shown by Billeter (1998), the basic excitation is caused by
galloping, because the pressure coe$cients C (0), C (0) and C (0) (pressure along the
N N N
gate bottom) increase steadily with z u /u [the function tan\(z u /u ) is the maximum
X X  X X 
angle of #ow incidence]. The vibration frequency f of the MIE-mechanism is only about
X
0)8 f (Figure 2, range C). For the small gate openings investigated (0)5(s/d(1), the
X
336 P. BILLETER AND T. STAUBLI

Figure 13. Transition to vertical MIE with simultaneous horizontal LEVS-excitation: vibration amplitudes
z /d (solid signs) and phases u(z !p ) (open signs); (, ) f /f +2)07, (, ) f /f +2)89, (, )
X X  V X V X
f /f +3)50.
V X

Figure 14. Transition to MIE, time series of vertical and horizontal amplitudes z (t) and x (t), < "13)0,
X V PX
f /f "2)89.
V X

hydrodynamic mean lift coe$cient is negative and almost proportional to the gate opening
(Thang & Naudascher 1986). Since the instantaneous gate opening is in-phase with the gate
de#ection, the mean lift force is in-phase with the gate de#ection, and it thus acts as negative
rigidity with regard to the vertical vibration. Consequently, the vibration frequency f drops
X
below f .
X
If the reduced velocities of both the horizontal and vertical modes happen to lie within
their excitation range, simultaneous vertical and horizontal vibrations z (t) and x (t) at
X V
di!erent frequencies can occur (Figure 14). As long as the vibration frequencies f and f are
V X
 FLOW-INDUCED MULTIPLE-MODE GATE VIBRATIONS 337

incommensurate, both the horizontal and the vertical vibration modes are caused by
independent excitation mechanisms (i.e., cross-#ow MIE and streamwise body-resonant
LEVS). Especially, the vertical vibration then behaves strongly nonlinearly with modula-
tions in the amplitude and the frequency.
Since the natural frequency of the vertical mode is most often smaller than the frequency
of the horizontal one for prototype gates, it was investigated whether vertical vibrations
associated with MIE could be excited by subharmonic modes of simultaneously occurring
horizontal vibrations. Figure 13 also shows the phases u between the vertical amplitude
X
z (t) and the pressure p acting on the gate lip bottom. Negative phases u refer to energy
X  X
transfer from the #ow to the body (the phase u is de"ned as p (t)"p sin(u t!u ). For
HL L L H HL
the -subharmonic mode of a horizontal vibration, lying between 8(< (11, the phases
 PX
are always positive. It can be concluded, therefore, that the MIE is not ampli"ed by
-subharmonic of the horizontal vibration. The onset of MIE occurs at < +11, as soon
 PX 
as the horizontal and the vertical vibration frequencies become incommensurate ( f /f '2).
V X

5. CONCLUSIONS
It was shown that #ow-induced vibrations of gates with two degrees of freedom (d.o.f.) are
caused by three dominant excitation mechanisms: cross-#ow and streamwise ILEV-excita-
tion (z- and x-ILEV), streamwise body-resonant LEVS (x-BR-LEVS) and cross-#ow MIE
(z-MIE). Characteristic properties of the di!erent excitation mechanisms were pointed out.
Depending on the frequency ratio of the two vibration modes and the type of excitation
involved, the mutual interaction of the two modes can a!ect both the vibration trajectory
(i.e., the size and the phase angle of the amplitudes) as well as the excitation range.

Figure 15. Characteristic vibration trajectories of under#own gates with one or two d.o.f. plotted as a function
of the reduced velocities < and < , respectively, and the frequency ratio f /f .
PX PV V X
338 P. BILLETER AND T. STAUBLI

A summary of these "ndings is given in Figure 15: the characteristic #ow-induced

vibration trajectories of under#own gates with one or two d.o.f. are shown as a function of
the reduced velocities < and < , respectively, and the frequency ratio f /f ( f /
PX PV V X V
f "0: 1-d.o.f. in x-direction, z-de#ection blocked; f /f PR: 1-d.o.f. in the z-direction,
X V X
x-de#ection blocked). For every vibration trajectory of the gate with 2-d.o.f., the maximum
amplitude of the dominant vibration mode is related to the reference amplitude of the
respective 1-d.o.f. gate vibration mode within the same excitation range; (reference ampli-
tudes of a gate with 1-d.o.f.: x is the maximum x-amplitude for x-ILEV excitation, z the
' '
maximum z-amplitude for z-ILEV exciation, x the maximum x-amplitude for x-BR-
*
LEVS excitation, and z the maximum x-amplitude for z-MIE excitation). Furthermore, it
+
is indicated whether the gate acts as a press-shut (PS) or a press-open device (PO).
As shown in Figure 15, the existence of a second d.o.f. rather e!ects the excitation range
and the vibration trajectory than the size of the maximum vibration amplitude. For
a frequency ratio close to one, however, cross-#ow ILEV and streamwise LEVS-excitation
coexist due to #uid dynamic coupling. Compared to gates with only 1 d.o.f., this mechanism
extends both the excitation range and the vibration amplitudes. More detailed results and
an extended analysis of the "ndings of the study presented here can be found in Billeter
(1998).

ACKNOWLEDGEMENT
Financial support, provided by the Swiss Federal Institute of Technology (ETHZ) and the
Laboratory of Hydraulics (VAW), is gratefully acknowledged.

REFERENCES
BILLETER, P. 1995 Flow-induced multiple-mode vibrations of a prototype weir gate. Proceedings
XX<I IAHR Congress, Vol. 1C, pp. 421}426, London, U.K.
BILLETER, P. 1998 StoK mungsinduzierte Schwingungen von SchuK tzen mit mehreren Freiheitsgraden.
<A= Mitteilung 125, Versuchsanstalt fuK r Wasserbau, Hydrololgie und Glaziologie der ETH
ZuK rich, ZuK rich (in German).
HARDWICK, J. D. 1974 Flow-induced vibrations of vertical-lift gate. ASCE Journal of Hydrautics
Division 100, 631}644.
ISHII, N. & KNISELY, C. W. 1992 Flow-induced vibration of shell-type long-span gates. Journal of Fluids
and Structures 6, 681}703.
ISHII, N. & KNISELY, C. W. 1990 Shear layer behaviour under a streamwise vibrating gate and its
signi"cance to gate vibration. JSME International Journal (Series III) 33, 131}138.
ISHII, N. & KNISELY, C. W. 1989 Shear layer dynamics and unsteady discharge beneath a long span gate
undergoing streamwise vibrations. Proceedings Hydrocomp, pp. 500}509. Dubrovnik: Elsevier.
JONGELING, T. H. G. 1988 Flow-induced self-excited in-#ow vibrations of gate plates. Journal of Fluids
and Structures 2, 541}566.
KANNE, S., NAUDASCHER, E. & WANG, Y. 1991 On the mechanism of self-excited vertical vibration of
under#ow gates. Proceedings International Conference on Flow-Induced <ibration, pp. 405}410,
Brighton. London: IMechE, Vol. 1991-6.
KOLKMAN, P. A. 1976 Flow-induced gate vibrations. Delft Hydraulics aboratory Publication, No. 164.
NAUDASCHER, E. & ROCKWELL, D. 1994 Flow-Induced <ibrations 2 An Engineering Guide; IAHR
Hydraulic Structures Design Manual Rotterdam: A.A. Balkema.
THANG, N. D. 1990 Gate vibrations due to unstable #ow separation. ASCE Journal of Hydrautic
Engineering 116, 342}361.
THANG, N. D. 1984 StroK mungsbedingte Schwingungen unterstroK mter SchuK tzen. Forschungsberichte
<DI, Reihe 4, No. 66, DuK sseldorf (in German).
THANG, N. D. & NAUDASCHER, E. 1986 Self-excited vibrations of vertical lift gates. IAHR Journal of
Hydrautic Research 24, 391}404.